Promptify入门学习资料 - 简化NLP任务的开源提示工程工具

Ray

Promptify简介

Promptify是一个开源的提示工程工具,旨在简化NLP任务的处理流程。它可以轻松生成适用于GPT、PaLM等流行生成模型的提示,帮助用户快速解决各种自然语言处理问题。

主要特点包括:

只需几行代码即可执行NER、文本分类等NLP任务,无需训练数据
支持添加one-shot、few-shot示例到提示中
提供结构化的Python对象输出,便于解析和过滤
支持自定义示例和样本
可在Hugging Face Hub上运行任何模型的推理
优化提示以减少OpenAI token成本(即将推出)

安装方法

推荐使用pip安装Promptify:

pip3 install promptify

或者从GitHub源码安装:

pip3 install git+https://github.com/promptslab/Promptify.git

快速上手

下面是一个使用Promptify进行医疗领域命名实体识别的简单示例:

from promptify import Prompter, OpenAI, Pipeline

sentence = """The patient is a 93-year-old female with a medical 
              history of chronic right hip pain, osteoporosis,
              hypertension, depression, and chronic atrial  
              fibrillation admitted for evaluation and management
              of severe nausea and vomiting and urinary tract
              infection"""

model = OpenAI(api_key) # 或使用HubModel()进行Huggingface推理
prompter = Prompter('ner.jinja') # 选择模板或提供自定义模板  
pipe = Pipeline(prompter, model)

result = pipe.fit(sentence, domain="medical", labels=None)

输出结果:

[
    {"E": "93-year-old", "T": "Age"},
    {"E": "chronic right hip pain", "T": "Medical Condition"},
    {"E": "osteoporosis", "T": "Medical Condition"},
    {"E": "hypertension", "T": "Medical Condition"},
    {"E": "depression", "T": "Medical Condition"},
    {"E": "chronic atrial fibrillation", "T": "Medical Condition"}, 
    {"E": "severe nausea and vomiting", "T": "Symptom"},
    {"E": "urinary tract infection", "T": "Medical Condition"},
    {"Branch": "Internal Medicine", "Group": "Geriatrics"},
]

NER示例